Converting Vocal, Bass, and Guitar MP3 to MIDI — What Actually Works

Why the instrument matters when the algorithm does not change

Our MP3 to MIDI converter does not have an instrument selector. It runs the same pitch-detection pipeline regardless of whether you feed it a vocal, a bass line, or a guitar riff. That is an honest description of how a monophonic browser detector works: it decodes the audio with the Web Audio API, runs autocorrelation frame by frame to find the dominant periodicity, maps that to the nearest MIDI note number, and writes a Standard MIDI File. There is no separate vocal model, bass model, or guitar model running behind the scenes.

What changes between instruments is the audio itself — the frequency range, the harmonic structure, the sustain envelope, and whether the sound is polyphonic. The detector responds differently to each of those, and that is what makes the instrument matter even when the code does not.

Vocals: the best-case instrument

A solo vocal recording is the input a monophonic detector handles best. The human voice produces a strong fundamental frequency with a clear periodic waveform. The autocorrelation function finds a clean peak at the period of that fundamental, and the note mapping is straightforward. If you hum a melody into your phone microphone and convert it, the detector can track the pitch contour.

Where vocals break down is when the recording is not isolated. If the vocal sits inside a full mix with drums, bass, and other instruments, the detector no longer sees a single pitch — it sees a composite waveform from multiple sources. The autocorrelation peak becomes ambiguous, and the output jumps between notes that do not match the melody. This is not a vocal-specific failure; it is the fundamental limitation of tracking a single pitch in polyphonic audio, which we explain in why MP3 to MIDI cannot hear chords.

Practical advice for vocals: use an a cappella recording, a hummed melody, or a vocal that has been isolated from the mix. The closer the input is to a solo voice and nothing else, the more usable the MIDI output will be. You can hum directly into your phone — our converter works on mobile, processes the audio in the browser, and never uploads the file to a server.

What to check after converting a vocal

• Open the detected-note list and compare it against the melody you sang. The note names should follow the contour you remember — up when you went up, down when you went down.
• Look for spurious short notes where the detector may have jittered between adjacent pitches. These are easy to delete in a piano roll.
• If the entire output is shifted by a consistent interval, the detector may have locked onto a formant or overtone. Transposing the whole track by that interval in a DAW fixes it.

Bass: octave errors are the main problem

Bass is the instrument most likely to produce octave errors in a monophonic detector. The reason is structural: a low bass note carries a fundamental together with a stack of harmonics, and those harmonics sit at frequencies where the autocorrelation also finds strong peaks. When the detector locks onto a harmonic instead of the fundamental, the note it writes is higher than the note that was played. Which note it writes depends on which harmonic won, so the error is not always the same and there is no correction that fixes every file. Listen to the result against the recording and correct it in a piano roll.

This is not a bug in the converter — it is a known property of autocorrelation-based pitch detection at low frequencies. The algorithm finds the period of the strongest repeating pattern in the waveform, and for a rich harmonic signal, that pattern can be the harmonic series rather than the fundamental alone. The lower the fundamental, the more harmonics fit within the analysis window, and the more candidates there are for the wrong answer.

You will know this happened when the MIDI output sounds like the right melody but in the wrong octave — the intervals between notes are correct, but everything is shifted up. The fix is straightforward: import the MIDI into a DAW, select all notes, and transpose down until the lowest note matches the lowest note you hear in the recording.

Preparing a bass recording

• Use a recording of the bass by itself if possible. A bass-and-drums mix is harder because the kick drum produces low-frequency energy that competes with the bass fundamental.
• If the bass is in a full mix, expect the detector to struggle. The output may jump between the bass note and whatever other instrument is loudest in the frame.
• After converting, always play back the MIDI through a bass sound and compare it to the original. If the contour is right but the octave is wrong, transpose. If the contour is wrong, the detector lost the bass entirely — try a more isolated section.

Guitar: single notes work, chords do not

A guitar playing a single-note melody line is a good candidate for MP3 to MIDI conversion. The plucked string produces a clear fundamental with a sharp attack and a decaying sustain — a waveform that autocorrelation tracks well. The note onset is unambiguous, and the pitch is stable during the sustain.

Strummed chords are where guitar conversion fails. A chord is multiple pitches sounding simultaneously, and a monophonic detector picks a single pitch from them — the loudest or the pitch with the strongest periodicity in the current analysis frame. The result is a single-note line that hops between chord tones unpredictably. It is not the root, it is not the melody, and it is not useful. This is the same limitation that applies to any polyphonic audio, and it is inherent to single-pitch tracking.

Distorted electric guitar adds another problem. High-gain distortion adds harmonic content that is extremely dense — the waveform is nearly sawtooth-like, and the autocorrelation finds multiple strong peaks at closely related periods. The detector may oscillate between the fundamental and a nearby harmonic, producing a jittery output with rapid note changes that do not exist in the recording.

What works for guitar

• Clean electric guitar playing single notes: good candidate. The signal is focused and the pitch is clear.
• Acoustic guitar playing a fingerpicked single-note line: workable, though the body resonance can add noise to the pitch estimate.
• Acoustic guitar strumming chords: not workable. The output will be a single note per strum, not the chord.
• Distorted rhythm guitar: expect instability. The detector may produce rapid spurious notes.
• Lead guitar solo in a mix: depends on how loud the solo is relative to the backing instruments. An isolated solo track converts far better than a solo buried in a mix.

A practical workflow for each instrument

Vocals

1. Record or obtain an isolated vocal — a cappella, hummed melody, or a vocal stem extracted from the mix.
2. Open the MP3 to MIDI converter and select the file. Conversion happens in your browser; the file is not uploaded.
3. Download the .mid and open the detected-note list. Verify the melodic contour matches what you sang.
4. Delete spurious short notes in a piano roll. If the whole track is shifted by a consistent interval, transpose.
5. Import into your DAW and assign an instrument. Optionally, run it through MIDI to WAV to hear the result.

Bass

1. Use a bass-only recording if possible. If the bass is in a mix, try to isolate a section where the bass is prominent and other instruments are quiet.
2. Convert the file and download the .mid.
3. Play the MIDI back through a bass instrument. Compare the melodic contour to the recording.
4. If the contour is correct but everything is in the wrong octave, select all notes in a piano roll and transpose down until the lowest note matches the recording.
5. If the contour is wrong — random jumps, notes that do not correspond to the bass line — the detector lost the bass. Try a different section or a cleaner recording.

Guitar

1. Identify a single-note passage — a riff, a solo, a fingerpicked line. Avoid strummed sections.
2. If the guitar is in a full mix, look for a section where the guitar is the loudest instrument, such as a solo break.
3. Convert the file and download the .mid.
4. Check the note list for jitter — rapid back-and-forth between adjacent pitches. If present, try a cleaner-tone recording or quantize the output in a DAW.
5. For distorted guitar, expect more cleanup. You may need to delete spurious notes and manually correct pitches that drifted.

Why there is no instrument selector

Some converters offer a dropdown: Piano, Guitar, Bass, Multi-instrument. What that dropdown does varies. In some cases it changes the frequency range the detector prioritizes. In others it switches to a different model — a polyphonic model for multi-instrument, a specialized low-frequency model for bass. Those are legitimate engineering choices, and they can improve results for specific inputs.

Our converter does not have that dropdown because it runs the same pipeline on every file: autocorrelation on the decoded audio, a single pitch at a time, written to a Standard MIDI File. We are not going to pretend that selecting "Bass" from a menu changes the algorithm when it does not. Instead, we tell you how the detector responds to each instrument type so you can prepare the file and interpret the output correctly.

The preparation matters more than a menu label. A clean isolated vocal will convert better than a full mix with "Vocal" selected. A single-note guitar line will convert better than a strummed chord with "Guitar" selected. The recording quality and the monophonic nature of the input are the dominant factors, and no instrument selector overrides that.

What every instrument shares

• A single pitch at a time. The detector tracks a single pitch per frame. Anything polyphonic — chords, double stops, harmonized vocals — gets reduced to a single note.
• Clean input beats clever settings. The biggest factor in output quality is how isolated the target instrument is in the recording.
• Inspect the output. Open the detected-note list, compare it to what you hear, and fix wrong notes in a piano roll. The converter gives you a starting point, not a finished transcription.
• No upload, no login. The file is decoded and processed entirely in your browser. Nothing is sent to a server, and no account is required. This is true for vocals, bass, guitar, and any other audio you convert.
• Works on mobile. You can record a hummed melody on your phone and convert it in the same browser session. No app installation needed.

Common mistakes by instrument

• Vocals: Feeding a full mix instead of an isolated vocal, then being surprised that the output does not match the melody. The detector cannot separate the voice from the instruments.
• Bass: Trusting the octave of the output without checking against the recording. Bass octave errors are common and fixable, but only if you notice them.
• Guitar: Expecting chord recognition from a monophonic detector. A strummed chord will not produce a MIDI chord — it will produce a single note, and not necessarily a useful result.
• All instruments: Converting a long, complex recording and expecting a clean result. Short, focused passages convert better because the detector has a clearer signal to work with.

Bottom line

Vocals, bass, and guitar each interact differently with a monophonic pitch detector, even though the algorithm is the same. Vocals are the most cooperative input when isolated. Bass is prone to octave errors that you fix by transposing. Guitar works for single-note lines and fails for chords. In each case, the recording quality and the isolation of the instrument matter more than any setting. Convert the cleanest single-line recording you can, inspect the note list, and plan to clean up the output in a DAW.